CHAPTER 36 / Model-based experiments

State feedback by pole placement

A position-sensed mass–spring–damper with designed state feedback and a Luenberger observer. Compare true and estimated states, noise and actuator limits.

THE SHORT ANSWER

Pole placement is a state-feedback design method that chooses the gain K so the closed-loop poles, the eigenvalues of A − BK, sit at locations you specify. For a two-state mass–spring–damper you build the desired characteristic polynomial from the chosen poles and match its coefficients to the closed-loop equation. The placed poles describe the unsaturated, exact-model linear loop only.

u=−Kx+Nr,det⁡(sI−(A−BK))=s2+2ζωns+ωn2u=-Kx+Nr,\qquad \det\bigl(sI-(A-BK)\bigr)=s^2+2\zeta\omega_n s+\omega_n^2
WHAT YOU WILL LEARN

Match the coefficients of a simple closed-loop equation to a desired pair of poles.

Controllability tells us whether pole placement is possible in a linear model. It does not choose the poles for us.

Review the previous lesson: Can you move it? Can you see it?
START WITH AN EXAMPLE

First, picture it

A position error tells a controller which way to push; velocity tells it when to brake. Weighting both signals lets us change how a mass returns to rest. We choose the desired motion through its poles, then calculate the weights.

THE IDEA

Build the idea step by step

Start with the mass-spring-damper equation. Feed back position and velocity with negative signs. Substituting that force law changes the coefficients of position and velocity in the original equation. Choose a desired polynomial from desired poles, then match coefficients. This is pole placement for this simple two-state example. Tracking a nonzero reference needs an additional reference term; regulating toward zero is enough for the worked example.

Read the equations

p¨+0.8p˙+4p=u,u=−k1p−k2p˙\ddot p+0.8\dot p+4p=u,\qquad u=-k_1p-k_2\dot p

Position feedback changes the stiffness-like coefficient. Velocity feedback changes the damping-like coefficient in this normalized unit-mass example.

s2+(0.8+k2)s+(4+k1)=0s^2+(0.8+k_2)s+(4+k_1)=0

This is the closed-loop characteristic equation after substitution.

What the symbols mean

pp
position; dots denote velocity and acceleration
uu
force input for a unit mass
k1,k2k_1,k_2
position and velocity feedback gains, with compatible units
ss
complex pole variable, in inverse seconds

The lab uses a continuous two-state mechanical model with position sensing and optional estimation. It includes a force limit. Predicted poles describe the unsaturated linear motion; they do not guarantee that a large starting error stays within that limit.

WORKED EXAMPLE

Put both poles at minus 2

  1. Choose a repeated pole at minus 2 inverse seconds. Multiply its factors to obtain the desired polynomial.

    (s+2)2=s2+4s+4(s+2)^2=s^2+4s+4
  2. Match the constant and linear coefficients to the closed-loop equation.

    4+k1=4,0.8+k2=44+k_1=4,\qquad0.8+k_2=4
  3. Choose zero position gain and velocity gain 3.2. This result assumes the stated model and unlimited force. A force limit can change the actual response.

    k1=0,k2=3.2k_1=0,\qquad k_2=3.2
More detail and model assumptions
A CLOSER LOOK

From a desired response to a polynomial

The lab's mass has position x and velocity v. Its open-loop equation is x″ + 0.8x′ + 4x = u. Position and velocity form the state vector; the input is force. Try a feedback law u = −k₁x − k₂v + Nr. Substitution gives x″ + (0.8+k₂)x′ + (4+k₁)x = Nr. The gains change the coefficients of the differential equation, so they change its poles.

Choose a natural frequency ωn and damping ratio ζ, giving the desired denominator s² + 2ζωn s + ωn². Matching coefficients yields k₁ = ωn²−4 and k₂ = 2ζωn−0.8. The lab computes these gains rather than asking you to guess them. For ωn = 2 and ζ = 0.7, K = [0, 2]. The resulting poles are approximately −1.4 ± j1.43. Doubling ωn at fixed ζ moves both poles proportionally farther from the origin in this nominal model.

A CLOSER LOOK

Reference scaling is not integral action

At a constant target with zero velocity, the nominal equilibrium satisfies (4+k₁)x = Nr. Choosing N = 4+k₁ = ωn² makes x = r in the unsaturated exact model. This prefilter is a model-based reference gain. It does not measure accumulated error and does not automatically reject an unknown constant load or a wrong spring coefficient. Integral augmentation would be a separate design.

The controllability test answers whether a chosen input can move every mode. It does not promise small gains or acceptable actuator effort. This two-state model is controllable, but a finite force limit can invalidate the nominal pole prediction. Inspect requested and applied force when increasing bandwidth. Pole placement chooses locations; LQR instead computes a gain from a quadratic state/input cost. Neither method removes the need to check constraints.

TRY IT

Connect the example to the lab

The worked example places a repeated pole at −2 with unlimited force. The connected experiment adds a ±10 N force limit, so compare the force it requests with the force it applies as you make the design faster.

Open the connected experiment
YOUR EXPERIMENT

Make a prediction before moving a control.

  1. PREDICTDoes doubling the desired natural frequency improve speed without any cost?
  2. CHANGE ONE THINGPin ωn = 2. Change only ωn to 4, then 6 rad/s. Compare position and requested/applied force. Keep damping and the initial state fixed.
  3. OBSERVEPin the baseline, then compare position with the requested and applied force at the same times.
  4. EXPLAINK is calculated from the desired second-order polynomial. Faster nominal poles increase the required feedback and reference gains. The ±10 N actuator can clip the request, so the actual response need not follow those poles. Pole placement describes the unsaturated, exact-model linear loop.
BE CAREFUL

The reference gain N = ωn² removes steady error only for the exact model. It is not integral action, so an unknown constant load or a wrong spring coefficient still leaves an error.

Check your understanding+

For the open-loop equation x″ + 0.8x′ + 4x = u, which gains place both closed-loop poles at −3?

The desired polynomial is (s+3)² = s² + 6s + 9. Matching it with s² + (0.8+k₂)s + (4+k₁) gives k₁ = 5 and k₂ = 5.2.
PRACTICE PROBLEMS

Solve these before opening the solutions

Work each problem on paper first. Open the answer to check yours, then the full solution if you need the steps, and mark how it went.

  1. For the lab's mass below, choose ωn = 3 rad/s and ζ = 0.6. Find k₁, k₂, the reference gain N and the closed-loop poles.

    x¨+0.8x˙+4x=u,u=−k1x−k2x˙+Nr\ddot x+0.8\dot x+4x=u,\qquad u=-k_1x-k_2\dot x+Nr
    Show answer

    Answerk₁ = 5, k₂ = 2.8 and N = 9; the poles are −1.8 ± j2.4.

    k1=5,k2=2.8,N=9,s=−1.8±j2.4k_1=5,\quad k_2=2.8,\quad N=9,\quad s=-1.8\pm j2.4
    Show full solution

    Solution

    1. Desired polynomial.

      s2+2ζωns+ωn2=s2+3.6s+9s^2+2\zeta\omega_ns+\omega_n^2=s^2+3.6s+9
    2. Substitute the feedback and match coefficients.

      s2+(0.8+k2)s+(4+k1) ⇒ k2=3.6−0.8=2.8,k1=9−4=5s^2+(0.8+k_2)s+(4+k_1)\ \Rightarrow\ k_2=3.6-0.8=2.8,\quad k_1=9-4=5
    3. Reference gain for zero steady error in the exact model.

      N=4+k1=ωn2=9N=4+k_1=\omega_n^2=9
    4. Closed-loop poles.

      s=−ζωn±jωn1−ζ2=−1.8±j2.4s=-\zeta\omega_n\pm j\omega_n\sqrt{1-\zeta^2}=-1.8\pm j2.4
  2. A balancing mechanism, linearized about its upright position, obeys the equation below and falls over without feedback. Find its open-loop poles, then the gains that place the closed-loop poles at −2 ± j2.

    x¨=2x+u,u=−k1x−k2x˙\ddot x=2x+u,\qquad u=-k_1x-k_2\dot x
    Show answer

    AnswerThe open-loop poles are ±√2 ≈ ±1.41, one of them unstable. k₁ = 10 and k₂ = 4.

    sOL=±2,k1=10,k2=4s_{\mathrm{OL}}=\pm\sqrt2,\qquad k_1=10,\quad k_2=4
    Show full solution

    Solution

    1. Open-loop characteristic equation.

      s2−2=0 ⇒ s=±2≈±1.41s^2-2=0\ \Rightarrow\ s=\pm\sqrt2\approx\pm1.41
    2. Substitute the feedback.

      s2+k2s+(k1−2)=0s^2+k_2s+(k_1-2)=0
    3. Match the desired polynomial.

      (s+2−j2)(s+2+j2)=s2+4s+8 ⇒ k2=4,k1−2=8⇒k1=10(s+2-j2)(s+2+j2)=s^2+4s+8\ \Rightarrow\ k_2=4,\quad k_1-2=8\Rightarrow k_1=10
    4. The position gain must first cancel the destabilizing +2 before it adds any stiffness, so every stabilizing design needs k₁ > 2. Controllability guarantees that the gains exist; it does not say the force they demand is available.

  3. The model below is in controllable canonical form. Find the open-loop poles and the gain K = [k₁ k₂ k₃] in u = −Kx that places the closed-loop poles at −2, −3 and −4.

    A=[010001−6−11−6],B=[001]A=\begin{bmatrix}0&1&0\\0&0&1\\-6&-11&-6\end{bmatrix},\qquad B=\begin{bmatrix}0\\0\\1\end{bmatrix}
    Show answer

    AnswerThe open-loop poles are −1, −2 and −3. K = [18 15 3].

    K=[18153]K=\begin{bmatrix}18&15&3\end{bmatrix}
    Show full solution

    Solution

    1. In this form the last row of A holds the characteristic coefficients with their signs reversed.

      det⁡(sI−A)=s3+6s2+11s+6=(s+1)(s+2)(s+3)\det(sI-A)=s^3+6s^2+11s+6=(s+1)(s+2)(s+3)
    2. With u = −Kx only the last row changes; it becomes [−6 − k₁, −11 − k₂, −6 − k₃].

      det⁡(sI−A+BK)=s3+(6+k3)s2+(11+k2)s+(6+k1)\det(sI-A+BK)=s^3+(6+k_3)s^2+(11+k_2)s+(6+k_1)
    3. Desired polynomial.

      (s+2)(s+3)(s+4)=s3+9s2+26s+24(s+2)(s+3)(s+4)=s^3+9s^2+26s+24
    4. Match coefficients.

      k1=24−6=18,k2=26−11=15,k3=9−6=3k_1=24-6=18,\qquad k_2=26-11=15,\qquad k_3=9-6=3
  4. Exam level. The lab mass ẍ + 0.8ẋ + 4x = u has an actuator limit of ±10 N. It starts at rest at x = 0 and the reference steps to r = 0.5 m. (a) For ζ = 0.7 and ωn = 6 rad/s, find k₁, k₂, N and the poles. (b) Find the requested force at the instant of the step and the steady force. (c) Repeat (b) for ωn = 2 rad/s with the same ζ. (d) What does this say about the 6 rad/s pole prediction?

    Show answer

    Answer(a) k₁ = 32, k₂ = 7.6, N = 36; poles −4.2 ± j4.28. (b) 18 N is requested at the step, so the actuator clips at 10 N; the steady force is 2 N. (c) With k₁ = 0, k₂ = 2 and N = 4 the initial request is 2 N, and the steady force is again 2 N. (d) The fast design saturates at the start, so its early motion does not follow the predicted poles.

    ωn=6: u(0+)=18 N>10 N;ωn=2: u(0+)=2 N;u∞=2 N\omega_n=6:\ u(0^+)=18\ \mathrm N>10\ \mathrm N;\qquad \omega_n=2:\ u(0^+)=2\ \mathrm N;\qquad u_\infty=2\ \mathrm N
    Show full solution

    Solution

    1. Gains from the lecture's formulas.

      k1=ωn2−4=32,k2=2ζωn−0.8=7.6,N=ωn2=36k_1=\omega_n^2-4=32,\quad k_2=2\zeta\omega_n-0.8=7.6,\quad N=\omega_n^2=36
    2. Poles.

      s=−0.7(6)±j61−0.49≈−4.2±j4.28s=-0.7(6)\pm j6\sqrt{1-0.49}\approx-4.2\pm j4.28
    3. At the instant of the step, x and ẋ are still zero, so only the reference term acts.

      u(0+)=Nr=36(0.5)=18 N>10 Nu(0^+)=Nr=36(0.5)=18\ \mathrm N>10\ \mathrm N
    4. At steady state x = r and ẋ = 0, so the force only balances the spring.

      u∞=−k1r+Nr=(N−k1)r=4r=2 Nu_\infty=-k_1r+Nr=(N-k_1)r=4r=2\ \mathrm N
    5. For ωn = 2: k₁ = 0, k₂ = 2(0.7)(2) − 0.8 = 2 and N = 4, so the initial request is well inside the limit and the pole prediction holds. The fast design spends its first moments clipped at 10 N and only approaches its nominal behaviour once the request falls below the limit.

      ωn=2:u(0+)=4(0.5)=2 N\omega_n=2:\quad u(0^+)=4(0.5)=2\ \mathrm N

Mark each exercise after you check your answer. Marks stay in this browser.

KEY TERMS

Terms this lecture teaches

All terms in the glossary →
QUESTIONS STUDENTS ASK

Frequently asked questions

How do I calculate the state-feedback gain K for pole placement?

Substitute u = −Kx into the state equation, write the characteristic polynomial of A − BK, and match its coefficients to the polynomial built from your desired poles. For the model ẍ + 0.8ẋ + 4x = u, that gives k₁ = ωn² − 4 and k₂ = 2ζωn − 0.8.

Why do faster closed-loop poles need larger gains?

Moving the poles farther from the origin raises the required stiffness and damping coefficients, so the gains grow with ωn. In the model ẍ + 0.8ẋ + 4x = u with ζ = 0.7, doubling ωn from 2 to 4 rad/s changes K from [0, 2] to [12, 4.8], and the requested force can then reach the actuator limit.

Can I place the closed-loop poles anywhere I want?

Only if the system is controllable from the chosen input; an uncontrollable mode keeps its eigenvalue whatever K you pick. Even when every pole can be moved, the theory does not promise small gains or acceptable actuator effort.

Does the reference gain N remove steady-state error like integral action?

No. Choosing N = ωn² makes the output equal the target only in the exact, unsaturated nominal model. N does not measure accumulated error, so an unknown constant load or a wrong spring coefficient still leaves an error; adding an integrator is a separate design.

Why the next lesson follows

Matching coefficients by hand works for two states. The next chapter turns it into a procedure for any order, Ackermann's formula, and adds an integrator so that the output reaches the target even when the model is slightly wrong.

YOUR PROGRESS

Finished the reading and the exercises? Mark the lecture as done to tick it on the syllabus. Progress stays in this browser.

EIGENROOM · https://eigenroom.com/theory/pole-placement · Updated 2026-11-15